Main site

Agriculture AI agent · Biosecurity

Biosecurity & Animal-Disease Event AI Agent

Watch premises signals, assemble the movement, permit and indemnity evidence an event needs, and prompt the call — the accredited veterinarian diagnoses and reports, and the state official decides the movement.

Typical delivery
4–6 weeks
Deployment
Your stack
Vet review
Premises-level
After launch
Agent Care

What this agent does

Raises the event, never the diagnosis

In

  1. 01

    A pen goes off feed, and the agent pulls the production, mortality and water records behind it.

  2. 02

    A tank result comes back, and the agent traces it to the herds and hauler routes that fed the load.

Reason

  1. 03

    A zone opens — screwworm has been live in Texas since June 2026 — and the agent maps the sites inside each ring.

  2. 04

    A load is planned, and the agent reads the movement terms in force for both states since the April 2026 update.

  3. 05

    A permit is needed, and the agent assembles the traceability, surveillance and biosecurity evidence for it.

Decide

  1. 06

    An audit clock runs down, and the agent flags the lapse the 2024 indemnity rule turns into lost eligibility.

  2. 07

    A flock is depopulated, and the agent builds the inventory, valuation and mortgage-consent packet.

Out

  1. 08

    An event closes, and the signals, the packet, the permit and the veterinarian's call stay on the premises.

  2. 09

    Execute write actions only inside the approval boundaries agreed during implementation.

Product statement

The agent watches and assembles; a veterinarian diagnoses and reports, and the state animal health official decides the movement.

Example workflow

One event, signal to permit

  1. 1Signal receivedProduction drop, mortality spike, bulk-tank result or a detection posted nearby
  2. 2Context assembledPremises records, movement history, zone status and audit validity, each with its named source
  3. 3Event packet draftedSignals, affected premises, permit evidence and confidence
  4. 4Controls appliedZone-ring checks, state-pair movement terms, audit-validity checks and confidence threshold

No human action required

Stages 1 to 4 run unaided, and nothing is reported or moved at any of them — the agent is assembling, and the veterinarian's lane opens at the confidence gate.

5DecisionBranches at the confidence threshold
High confidence

Goes to the herd veterinarian to review.

Low confidence

Adds a biosecurity coordinator read first.

Veterinarian review

The packet is held with its signals, its zone status and the confidence.

Approve · Edit · Hold for a farm visit

Approved — released to report

  1. 6Herd and premises records updatedOnly where write access and approval policy allow it
  2. 7Outcome evaluatedTime to report, permit turnaround, audit lapses and corrections made after the movement

Edits

Every veterinarian edit is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required

8 items

  • Diagnosing the disease or confirming a case.
    • Making the report to the state animal health official.
    • Granting or issuing a movement permit.
    • Signing a certificate of veterinary inspection.

  • Automation boundary

    Agent acts unaided

    • ✓Watch production, mortality and bulk-tank signals.
    • ✓Assemble the evidence a permit application needs.
    • ✓Track zone rings, premises status and audit validity for the coordinator.
    • ✓Prompt the call, and hold the packet for the herd veterinarian to work from.

    Any write happens inside the boundaries agreed at implementation, never ahead of a permit.

    • Signing the appraisal and indemnity claim form.
    • Setting zone boundaries or premises designations.
    • Passing or waiving an APHIS biosecurity audit.
  • Changes to zone, movement or reporting rules.

Example output

One event packet, annotated

Everything the agent assembles is attached to the record it was drawn from.

Event-packet output · single premises

Illustrative example
Premises
Layer house, contract grower
Signal raised
Daily mortality has climbed for two days and water intake is down in one house
Zone status
Buffer zone
Source of record
Barn controller and daily log
Confidence
91%
Attribution
Named veterinarian and accreditation ID

As received

Taken from the barn controller and the daily log — nothing on this side is written by the agent.

Signals used

  • Daily mortality log
  • Water and feed intake
  • Nearby detection notice

Why this was raised

Birds that die before depopulation are not indemnified.

What the score decides

Below the configured threshold the packet picks up a coordinator read first.

Action

ApproveEditHold for a farm visit

Value

Where AI adds value

The same four claims, placed at the point in the workflow where each one applies.

Where the value lands

Value 01 – 04

Every premises

From the premises records

03Assembly

Assemble from the record

Draw on the premises records and the zone, movement and audit rules configured for the operation.

01Approved path

Reporting beats diagnosing

Routine surveillance and permit evidence arrive already gathered.

02Human review

Send the veterinarian to the risk

Threshold crossings and lapsing audits are marked, so the veterinarian's read starts where the risk sits.

04

Build an evidence trail

The event, the premises it was raised on and the veterinarian who called it stay on the record.

Integrations

Typical integrations

Five system groups connect to the same agent. Which of them are in scope is decided in discovery.

  • Herd & flock records

    • DairyComp 305
    • PCDART
    • Poultry flock systems

  • Milk & lab results

    • NAHLN lab portals
    • Co-op bulk-tank results

  • Premises & movement

    • Premises ID
    • CVI records
    • State permit portals

Agent

Biosecurity & disease events

  • Reads the signals
  • Assembles the packet
  • Holds for the vet

  • Environment & alerts

    • Barn controllers
    • sensors
    • APHIS detection feeds

  • Observability & evaluation

    • OpenTelemetry
    • Langfuse
    • Supported monitoring/evaluation sources

Integration availability depends on the client's existing systems and API access.

Agent controls

Six layers between the model and the premises

Six controls, each one folded inside the last. What comes through them is named in the map below.

L6 · Outermost — last line of defence

Inward → L1 · closest to the model

L6Rollback / safe modeFall back to signal watching when evaluation or production signals degrade.Roll back
L5Version monitoringTrack model, prompt, zone-rule and state-movement-term changes.Track
L4TraceabilityRecord the signals, the packet, the permit evidence and the veterinarian's call.Record
L3Veterinarian reviewHold packets for the named veterinarian; it governs release, not whether the signal was read right.Gate
L2Policy guardrailsTest packets against the configured zone, movement and audit rules; a failure returns the packet.Restrict
L1Confidence thresholdsRoute low-confidence packets to a coordinator read before the veterinarian sees them.Require review
Model corePacket assembled — signals, affected premises, permit evidence and confidence
L1 – L2
Test whether a packet may stand
L3
Puts the call in a veterinarian's hands
L4 – L5
Keep the event and the premises behind it
L6
Holds the movement at unapproved when signals degrade

How Nestack evaluates it

Evaluate the event workflow — not only the finished packet.

Coverage runs the whole depth of the workflow, and every layer is cut by slice.

Surface — the packet the veterinarian reads

  1. E1Final-output evaluationDid every signal in the packet trace back to the record it came from?
  2. E2Step-level evaluationDid the agent use the right premises, zone status and the movement terms in force that day?
  3. E3Tool evaluationDid it read and write the correct premises and the correct field?
  4. E4Confidence calibrationDo low-confidence packets actually attract more veterinarian edits?
  5. E5Slice evaluationHow does performance change across specific premises types?
  6. E6Business outcomeHow many packets needed a veterinarian edit or a correction after the event closed?

Floor — the outcome the premises answers for

Failure modes

Where each failure originates in the agent

Seven failure modes, placed at the stage each one originates.

Agent lifecycle

Direction of processing →

  1. 01 · Retrieval

    1 mode

    • QB-03

      Stale state status

      Movement terms read from a list that changed after the load was planned.

    Stage gathers

    Premises records, zone status, movement terms and audits

  2. 02 · Reasoning

    2 modes

    • QB-04

      Sampling window missed

      Milk samples predate the movement by more than seven days.

    • QB-06

      Audit read as current

      A coordinator or barn change voids it and nothing flags.

    Stage proposes

    Affected premises, permit evidence, confidence

  3. 03 · Tool / write

    2 modes

    • QB-02

      Permit not yet granted

      Submission is treated as issuance and the load leaves.

    • QB-05

      Split load uncovered

      A tested group is split and each lot is not re-checked.

    Stage writes

    Only where write access and approval policy allow it

  4. 04 · Output

    1 mode

    • QB-01

      Case queued, never called

      The packet opens and the immediate call is never placed.

    Stage returns

    The packet the veterinarian reads and acts on

  5. 05 · Change / Version

    1 mode

    • QB-07

      Silent rule drift

      A zone or movement-term update widens what the agent will clear.

    Stage tracks

    Model, prompt, zone rules and movement terms

  • Sev-1 · a movement runs without a permit
  • Sev-2 · a wrong signal reaches the packet
  • Sev-3 · records degrade, packet goes to review
See the complete Agriculture failure-mode catalogue

Affected slices

A whole-system figure can hide one bad site

A whole-system reporting time can look acceptable while a few premises cohorts absorb most of the rework that delays it. Nestack reports the packet-rework rate by slice, not only in total.

Slice performance — reported separately, not only in aggregate

Illustrative example
SliceFailure rateLiftLift vs. thresholdStatus
Contract-grower premises8.9%3.8×Review
Interstate dairy movements5.2%2.2×Review
Buffer-zone premises3.7%1.6×Watch
Single-site owned premises2.1%0.9×Normal

Bar: packet-rework-rate lift vs. the single-site baseline · scale 0–4.0× · tick marks the 2.0× review threshold

2 of 4 slices over threshold

Evidence-linked improvement

Every cycle closes with a new case

A cycle closes when the late report is a case the next release has to catch. That suite is what the next event worked is measured against.

Improvement cycle · five stages

Switchback — the path turns at Improve and returns at Learn

  1. 01

    Detect

    Packet-rework rate rises in a premises slice.

  2. 02

    Diagnose

    The four days between the first sick pen and the call are walked back until the cause narrows to one.

  3. 03

    Improve

    The correction is numbered and the events that produced it are pinned to the note.

  4. 04

    Verify

    One movement case still failing is enough to keep the release on the bench.

  5. 05

    Learn

    The case is kept, and the zone rules are updated in the same change.

  6. Learn → Detect

    The return edge. The next event is worked against a suite one case longer.

Typical build scope

Twelve workstreams across six weeks

The build scope read against the delivery timeline. Week structure follows the six-week plan — discovery, sources, event workflow, evaluation, integration, then production validation and handover.

Workstream

Week 1Week 2Week 3Week 4Week 5Week 6

  1. 01Disease-event workflow and premises boundary definition.Week 1
  2. 02Herd, flock and lab source assessment.Week 1, Week 2
  3. 03Zone-ring, movement-term and audit-validity mapping.Week 1, Week 2
  4. 04Premises-record ingestion and mapping.Week 2
  5. 05Signal logic and record binding.Week 2, Week 3
  6. 06Confidence scoring and event routing.Week 3
  7. 07Veterinarian review workflow.Week 3, Week 4
  8. 08Herd-record and permit-portal integration.Week 5
  9. 09Movement and reporting cases.Week 4, Week 5
  10. 10Guardrails and permit controls.Week 4
  11. 11Premises-trail instrumentation.Week 5, Week 6
  12. 12Deployment, documentation and Agent Care handover.Week 6

12 workstreams · 6 weeks · bar shows the weeks a workstream is active — several run in parallel

Final scope and sequence confirmed in discovery

Engagement tiers

What each tier includes

Rows are the capabilities named in each tier's scope. Higher tiers include everything below them.

Capability✓ in scope · — not at this tierPilotOne premises, one speciesProductionProduction herd systemsAdvancedMultiple sites / states
Introduced at Pilot
Signal watching to your thresholds✓✓✓
Veterinarian review✓✓✓
Reporting-speed baseline✓✓✓
Introduced at Production
Reporting by premises—✓✓
Review workflow in your systems—✓✓
Approved write-back—✓✓
Herd-record integration—✓✓
Introduced at Advanced
Multi-state movement rules——✓
Multi-stage event approvals——✓
High premises count——✓
Multi-site movement controls——✓
Build priceFrom $5,000From $8,000Custom quote
Final build price
Confirmed after discovery based on integrations, workflow complexity, premises count, review controls and deployment requirements.
Separate from build
Build pricing is separate from recurring Agent Care, which covers managed monitoring, evaluations, incidents and verified improvements after launch.

What we need from you

What you bring, and what we build with it

Each input maps to a piece of build scope and a week in the delivery timeline.

  1. 01Your premises list and species mix

    Premises-record ingestion and mapping

    Week 1

  2. 02Representative past events and movements

    Signal baseline, threshold extraction and record binding

    Week 2

  3. 03Your zone, movement and audit rule sources

    Zone, movement and audit rule mapping

    Week 1

  4. 04Access to relevant APIs, feeds or exports

    Herd, flock and lab assessment, then integration setup

    Week 2

  5. 05Movements you would not want permitted

    Outbreak cases and the evaluation suite

    Week 4

  6. 06What no movement may proceed without

    Confidence scoring, event routing, guardrails and permit controls

    Week 3

  7. 07Named veterinarians to review packets

    Veterinarian review workflow, then pilot and production validation

    Weeks 5–6

Nothing else is required

Deployment, documentation and Agent Care handover are ours.

Delivery timeline

Four phases across six weeks

Each band lands where the work lands, which puts evaluation and the first live events in one week.

Phase

W1

W2

W3

W4

W5

W6

  1. DiscoveryW1
  2. BuildW2 – W3
  3. EvaluateW4 – W5
  4. Pilot & LaunchW5 – W6

Week focus

W1Event workflow discovery, rule mapping and the automation boundary

W2Herd, flock and lab integration and the signal baseline

W3Event workflow, confidence logic and veterinarian-review controls

W4Evaluation suite, movement-rule checks and failure-mode testing

W5Permit-portal integration, pilot events and targeted corrections

W6One surveillance cycle run under the herd veterinarian, then handover

Reading the band
A bar covers the weeks its work is named in, and nothing else. The week 5 overlap is real, not padding.
At the end of W6
Validation closes on live events, and Agent Care takes over monitoring.
Duration
Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Agriculture AI agent

Build a biosecurity agent around your premises and your veterinarian.

Show us your premises list, your movement terms and who you call. What changes first? Time-to-report — and on a poultry claim that is the number deciding what is recoverable.

Typical delivery
4–6 weeks
Deployment
Your herd and flock records
Vet review
Premises-level
After launch
Agent Care available

Nestack Agents · Biosecurity & disease events

AGT-AGR-16 · Agent Care available after launch

Build a biosecurity agent

Four questions so the call starts from your actual setup — then pick a time that suits you.

About you
Your premises

No obligation. We’ll map the workflow, name the automation boundary and tell you if an agent is the wrong answer.